The relationship between line bisection performance and emotion processing: Where do you draw the line?

ABSTRACT

A recent study demonstrated that higher accuracy on a line bisection task related to greater ratings of evocative impact from paintings. The authors suggested that line bisection accuracy may act as a “barometer” for both visuospatial and emotion processing, likely as a function of overlapping neural correlates in the right temporoparietal region. We suggest and test an alternative explanation: that visuospatial bias interacted with asymmetries in the paintings and the rating scales to produce the apparent relationship between emotion and visuospatial functions. In the present study, using both visual-analogue and numeric rating scales, the relationship between line bisection performance and ratings of paintings (evocative impact, aesthetics, novelty, technique, and closure) was examined in a young adult sample. We demonstrate that left-hand line bisection bias direction, not line bisection accuracy, is related to most ratings, and that line bisection bias interacts with stimulus orientation (non-mirrored/mirrored) and rating scale direction (ascending/descending) in such a way that can explain the results of the previous study. We conclude that the line bisection task appears to be a sensitive measure of visuospatial attentional biases, which can influence ratings of asymmetrical paintings, and may affect how individuals perceive stimuli in their environment.

KEYWORDS:

• Line bisection
• visuospatial attention
• emotion processing
• temporoparietal
• art

Acknowledgements

The authors gratefully thank Chris Oriet for his feedback on the manuscript and revision, and Dennis Alfano and Katherine Robinson for their feedback on an earlier version of the manuscript. We also thank Brendan Demyen, Denis Gavigan, Jamie Oakenfold, and Mark Adkins for their assistance with data scoring and entry, along with members of C5 at the University of Regina for their comments on the design and early results of this project.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹There was no evidence of multicollinearity, as no tolerance values were less than .377 (Cohen, Cohen, West, & Aiken, 2003). Three possible outliers were identified using Studentized Deleted Residuals, and Shapiro–Wilk's tests indicated that the Studentized residuals were normally distributed (all ps > .05). However, these possible outliers were not overly unusual, as Cook's distances were all smaller than 1, indicating that there were no influential cases (Cook & Weisberg, 1982), and leverage points were fairly close to the suggested cutoff, indicating no unusual combination of the independent variables. In addition, the assumption of homoscedasticity was reached, based on visual inspection of studentized residuals plotted against predicted values for mirrored and non-mirrored stimuli. Taken together, there is sufficient evidence that our data meet the necessary assumptions.

²There was no evidence of multicollinearity, nor were there any outlying cases according to the Studentized Deleted Residuals. Shapiro–Wilk's tests indicated that the Studentized residuals were normally distributed (all ps > .08). Cook's distances and leverage values all fell within their expected ranges. Homoscedasticity was reached based on visual inspection of Studentized residuals plotted against predicted values for ascending and descending scale type. Altogether, our data met the necessary assumptions for this analysis.

³The paintings were converted into 1-bit black and white images using GNU Image Manipulation Program (GIMP), and the percentage of white pixels in the left and right halves of each painting was calculated and then compared, using a paired samples t-test. The left halves had significantly more white pixels than the right halves, t(7) = 4.01, SEM = 4.33, p = .005.

⁴This was further supported by supplementary analyses using z difference scores between LLB and RLB correlations (Table 1). Table A (see online supplemental file at http://dx.doi.org/10.1080/1357650X.2015.1134564) demonstrates that attribute rating correlations with LLB and RLB biases do significantly differ from each other, whereas Table B (see online supplemental file at http://dx.doi.org/10.1080/1357650X.2015.1134564) demonstrates that correlations between line bisection biases and attribute ratings made on the two scale formats (visual-analogue and numeric) do not differ from each other.

Additional information

Funding

This research was supported by an NSERC PGSD granted to the first author, and a Canada Foundation for Innovation grant awarded to the second author.